tau_parameters.h

Go to the documentation of this file.

#ifndef PARAMETERS_H_
#define PARAMETERS_H_

#include <complex>
#include "ap_int.h"
#include "ap_fixed.h"
#include "L1Trigger/Phase2L1ParticleFlow/interface/common/nnet_layer.h"
#include "L1Trigger/Phase2L1ParticleFlow/interface/common/nnet_activation.h"
#include "L1Trigger/Phase2L1ParticleFlow/interface/common/nnet_common.h"

//hls-fpga-machine-learning insert numbers
#define N_INPUTS 80
#define N_LAYER_1 25
#define N_LAYER_2 10
#define N_LAYER_3 10
#define N_OUTPUTS 1

//hls-fpga-machine-learning insert layer-precision

typedef ap_fixed<24, 12> input2_t;
typedef ap_fixed<16, 8> input_t;
typedef ap_fixed<16, 8> layer1_t;
typedef ap_fixed<16, 8> layer2_t;
typedef ap_fixed<16, 8> layer3_t;
typedef ap_fixed<16, 8> result_t;
typedef ap_fixed<16, 8> accum_default_t;
typedef ap_fixed<16, 8> weight_default_t;
typedef ap_fixed<16, 8> bias_default_t;

//hls-fpga-machine-learning insert layer-config
struct config1 : nnet::layer_config {
  static const unsigned n_in = N_INPUTS;
  static const unsigned n_out = N_LAYER_1;
  static const unsigned io_type = nnet::io_parallel;
  static const unsigned reuse_factor = 1;
  //static const unsigned reuse_factor = 6;
  static const unsigned n_zeros = 0;
  static const bool store_weights_in_bram = false;
  typedef accum_default_t accum_t;
  typedef bias_default_t bias_t;
  typedef weight_default_t weight_t;
};
struct relu_config1 : nnet::activ_config {
  static const unsigned n_in = N_LAYER_1;
  static const unsigned table_size = 1024;
  static const unsigned io_type = nnet::io_parallel;
};
struct config2 : nnet::layer_config {
  static const unsigned n_in = N_LAYER_1;
  static const unsigned n_out = N_LAYER_2;
  static const unsigned io_type = nnet::io_parallel;
  static const unsigned reuse_factor = 1;
  //static const unsigned reuse_factor = 6;
  static const unsigned n_zeros = 0;
  static const bool store_weights_in_bram = false;
  typedef accum_default_t accum_t;
  typedef bias_default_t bias_t;
  typedef weight_default_t weight_t;
};
struct relu_config2 : nnet::activ_config {
  static const unsigned n_in = N_LAYER_2;
  static const unsigned table_size = 1024;
  static const unsigned io_type = nnet::io_parallel;
};
struct config3 : nnet::layer_config {
  static const unsigned n_in = N_LAYER_2;
  static const unsigned n_out = N_LAYER_3;
  static const unsigned io_type = nnet::io_parallel;
  static const unsigned reuse_factor = 1;
  //static const unsigned reuse_factor = 6;
  static const unsigned n_zeros = 0;
  static const bool store_weights_in_bram = false;
  typedef accum_default_t accum_t;
  typedef bias_default_t bias_t;
  typedef weight_default_t weight_t;
};
struct relu_config3 : nnet::activ_config {
  static const unsigned n_in = N_LAYER_3;
  static const unsigned table_size = 1024;
  static const unsigned io_type = nnet::io_parallel;
};
struct config4 : nnet::layer_config {
  static const unsigned n_in = N_LAYER_3;
  static const unsigned n_out = N_OUTPUTS;
  static const unsigned io_type = nnet::io_parallel;
  static const unsigned reuse_factor = 1;
  //static const unsigned reuse_factor = 6;
  static const unsigned n_zeros = 0;
  static const bool store_weights_in_bram = false;
  typedef accum_default_t accum_t;
  typedef bias_default_t bias_t;
  typedef weight_default_t weight_t;
};
struct sigmoid_config4 : nnet::activ_config {
  static const unsigned n_in = N_OUTPUTS;
  static const unsigned table_size = 1024;
  static const unsigned io_type = nnet::io_parallel;
};

#endif